French Publishers: ChatGPT Cited More Often Than Claude

    New data from French publishers has revealed significant differences in how artificial intelligence systems cite their sources. While ChatGPT dominates with a 32% share, competing models like Claude or Grok refer to original content significantly less.

    MediaInsideMediaInside Editorial Team5 min read
    French Publishers: ChatGPT Cited More Often Than Claude
    Share:

    French research firm Occurrence has published the first comprehensive data on "Citation Share," the proportion of citations that artificial intelligence attributes to specific sources. The analysis, presented at the PR Lab 2026 conference in Paris in March, shows fundamental differences in how various language models handle brand and media content. OpenAI's ChatGPT holds the strongest position, citing corporate sources in 32% of cases.

    The Chasm Between ChatGPT and Competitors

    The study results confirm that OpenAI's market dominance is not only in user numbers but also in the extent to which the tool refers to primary information sources. While ChatGPT leads with a significant margin, its main competitors lag:

    • Gemini (Google): 26 %
    • Mistral: 22 %
    • Perplexity: 19 %
    • Claude (Anthropic): 17 %
    • Grok (xAI): 13 %

    A particularly interesting finding is the low citation share for the Claude model. Although considered technologically advanced by experts, it cites company websites almost twice less than ChatGPT. According to experts, this is due to Anthropic's more conservative approach to source diversity and a later market entry, which affected the speed of adoption among professionals.

    New Challenges for the Media Market

    In addition to technical statistics, the report highlights the growing tension between publishers and technology giants. Pierre Petillault, director of l’Alliance de la Presse d’Information Générale, which represents 300 French publishers, criticized AI companies.

    “Large language models have used French press content on a massive scale without authorization or any compensation. Obtaining a public admission from them that this data collection occurred before launch is very difficult.”

    Occurrence's analysis further suggests an emerging era of Generative Engine Optimization (GEO). It was found that 30 to 40% of the sources from which AI draws are channels directly influenceable by brands or publishers. For communication teams, this means that a strategy based solely on one model, most often ChatGPT, may not work on other platforms.

    AI Analysis

    AI

    Automatically generated commentary • Does not replace expert analysis

    Let AI analyse this article — market impact, comparison and historical context.

    Early Signal Detector

    AI

    Captures signals before the mainstream — inspired by Dataminr

    Reveal early signals and trends the mainstream has not yet captured.

    Context Connector

    AI

    Historical parallels, similar cases and hidden patterns

    Discover what connects to this article — historical parallels, similar cases and cross-industry insights.

    MediaInside
    MediaInside Editorial Team

    Editorial Team

    Under the brand MediaInside Editorial Team, texts are published that are the result of collective editorial work – summaries, short news, overviews, and comments, where the entire editorial team, not one specific author, is responsible for the content.